Abstract
In this article, we evaluate the impact of positional and temporal inaccuracies on the mapping and detection of potential outbreaks of dengue fever in Cali, an urban environment of Colombia. Positional uncertainties in input data are determined by comparison between coordinates following an automated geocoding process and those extracted from on-field GPS measurements. Temporal uncertainties are modeled around the incubation period for dengue fever. To test the robustness of disease intensities in space and time when accounting for the potential space-time error, each dengue case is perturbed using Monte Carlo simulations. A space-time kernel density estimation (STKDE) is conducted on both perturbed and observed sets of dengue cases. To reduce the computational effort, we use a parallel spatial computing solution. The results are visualized in a 3D framework, which facilitates the discovery of new, significant space-time patterns and shapes of dengue outbreaks while enhancing our understanding of complex and uncertain dynamics of vector-borne diseases.
Notes
1. We used a Trimble Juno SB GPS unit (2–5 m spatial accuracy) and a Trimble Geo XH GPS unit (2–3 cm spatial accuracy).
2. Neighborhoods where GPS data were collected: El Lido, Nueva Tequendama, Pampalinda, Urbanización Nueva Tequendama, Prados del Norte, Vipasa, and La Merced. The seven neighborhoods differ on their infrastructure condition.
3. STKDE using small spatial and temporal values will generally result in very spiky cluster, while larger bandwidth may result in ‘oversmoothing’ (Bailey and Gatrell Citation1995). Our approach strikes a balance between those two extremes.